Spaced Repetition

Spaced repetition is a learning technique that schedules each review of a piece of material at increasing intervals — one day, then several days, then weeks — timed to land just before you would otherwise forget it.

It is built on two of the oldest and most replicated findings in cognitive science. Hermann Ebbinghaus's 1885 experiments mapped the forgetting curve: memory for new material decays steeply — much of it lost within days — unless reviewed. And the spacing effect, documented in the same work and confirmed across a century of studies, shows that reviews distributed over time produce far stronger retention than the same total study time massed into one session. Cramming and spacing can look identical on tomorrow's test; a month later, they are not remotely comparable.

How it works

Every successful review flattens the forgetting curve, so the next review can wait longer. A typical trajectory for one item: learn it today, review after 1 day, then 6 days, then 2 weeks, then 5 weeks, then months. Miss the timing in either direction and you lose efficiency — review too soon and it is wasted repetition of something you still knew; too late and you are relearning from scratch rather than reinforcing.

The struggle is a feature. Retrieval practice — forcing yourself to recall the answer rather than rereading it — is itself one of the strongest known learning interventions (the testing effect), and spacing makes each retrieval effortful enough to count. Because tracking optimal intervals for hundreds of items is impossible by hand, software does the bookkeeping: algorithms like SM-2 compute each item's next due date from your review history.

Why it matters

For engineers, the failure mode spaced repetition fixes is universal: you learned how TLS handshakes work, or what a B-tree index does, passed the moment, and six months later the knowledge is gone when a production incident needs it. Breadth of knowledge — the kind that lets you connect a symptom in one system to a cause in another — only exists if it is retained, and retention does not happen by exposure. It happens by scheduled retrieval.

A worked example

Monday: you learn that TCP connection setup is a three-way handshake — SYN, SYN-ACK, ACK.

Tuesday (day 1): prompted to recall it. Effortful but successful.

Next Monday (+6): recalled faster. Interval extends.

Two weeks later: still there. Interval extends to a month.

Total investment: roughly four minutes across a month. Compare the alternative — read it once, feel fluent, and rediscover during an incident review that fluency evaporated in a week. Four minutes, distributed, bought you a durable fact; forty minutes, massed, would not have.

How miatz uses it

Spaced repetition is the backbone of the miatz daily training loop. Every concept across the 28 competencies enters an SM-2 scheduled review queue; each morning's session serves exactly what is due, and Saturday reviews consolidate the week. Breadth across a T-shaped engineer's whole base is only honest if it survives the forgetting curve — so miatz schedules the fight against it, every day.